UAV-Based Warehouse Management Using Multi-Agent RL

Abstract

This chapter aims to investigate the applications of reinforcement learning (RL) and Multi-Agent RL (MARL) in UAV networks in warehouse management. Different applications of UAVs in warehousing and different applications of RL in UAV networks are reviewed. Currently, most research in this area relies on single-agent RL approaches. Transitioning from single-agent RL to MARL offers the opportunity can potentially achieve higher levels of optimization, scalability, and adaptability in warehouse management using UAV networks. However, the application of multi-agent approaches in UAV-based warehouse management is still in its early stages, and may introduce new challenges. Thus, this chapter specifically focuses on the challenges and solutions associated with adopting MARL in the context of UAV-based warehouse management tasks. This includes addressing challenges such as non-stationary, partial observability, credit assignment, scalability, and task allocation. The authors highlight challenges and present some of the widely-used approaches to address these challenges.